Probabilistic risk assessment

Category: Risk identification & assessment · Reviewed by Al Jabbar, Broker · Specialist Risks · Last reviewed

Probabilistic risk assessment

Probabilistic risk assessment (PRA) is a quantitative methodology that uses probability theory and statistical modelling to estimate the likelihood and consequence of adverse events. It was pioneered in the US nuclear industry following the Reactor Safety Study (WASH-1400, 1975) and has since been adopted across aerospace, oil and gas, and finance.

Three core questions

The classical Kaplan-Garrick formulation poses three questions:

1. What can go wrong? (Identification)
2. How likely is it? (Frequency)
3. What are the consequences? (Severity)

The answers form a risk triplet — scenario, probability, consequence — and the set of all triplets defines the risk profile.

Techniques used

• Fault trees and event trees.
• Markov chains for systems with time-dependent state changes.
• Bayesian belief networks.
• Monte Carlo simulation.
• Extreme value theory for the tail.

PRA in Solvency II

The 99.5% one-year VaR calibration of the SCR is a probabilistic risk assessment of the firm. The ORSA is the firm’s own articulation of its PRA.

References

• Kaplan, S. and Garrick, B.J. (1981). On the Quantitative Definition of Risk. Risk Analysis 1(1).
• US NRC (1975). Reactor Safety Study (WASH-1400).
• IFoA practical guides on risk and capital.

Cross-references

• Monte Carlo simulation
• Quantitative risk assessment

Maintained by Matt Bartlett, Director, Apex Insurance Brokers Limited. FCA FRN 724952. Companies House 07014570.